func TestRefreshPodIsGone(t *testing.T) {
uuid, err := types.NewUUID("aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaaa")
if err != nil {
panic(err)
}
tmpDir, err := ioutil.TempDir("", "")
if err != nil {
panic(err)
}
defer os.RemoveAll(tmpDir)
podPath := filepath.Join(embryoDir(tmpDir), uuid.String())
os.MkdirAll(podPath, 0777)
p, err := getPod(tmpDir, uuid)
if err != nil {
t.Fatalf("unable to get pod: %v", err)
}
os.RemoveAll(tmpDir)
err = p.refreshState()
if err != nil {
t.Fatalf("error refreshing state: %v", err)
}
pstate := podToStates(p)
expected := states{isGone: true}
if !reflect.DeepEqual(expected, pstate) {
t.Errorf("expected %+v == %+v after refrshState", expected, pstate)
}
}
func handlePodVerify(w http.ResponseWriter, r *http.Request) {
defer r.Body.Close()
uuid, err := types.NewUUID(r.FormValue("uuid"))
if err != nil {
w.WriteHeader(http.StatusBadRequest)
fmt.Fprintf(w, "uuid field missing or malformed: %v", err)
return
}
content := r.FormValue("content")
if content == "" {
w.WriteHeader(http.StatusBadRequest)
fmt.Fprintf(w, "content field missing")
return
}
sig, err := base64.StdEncoding.DecodeString(r.FormValue("signature"))
if err != nil {
w.WriteHeader(http.StatusBadRequest)
fmt.Fprintf(w, "signature field missing or corrupt: %v", err)
return
}
h := hmac.New(sha512.New, hmacKey[:])
h.Write((*uuid)[:])
h.Write([]byte(content))
if hmac.Equal(sig, h.Sum(nil)) {
w.WriteHeader(http.StatusOK)
} else {
w.WriteHeader(http.StatusForbidden)
}
}
func main() {
flag.Parse()
log, diag, _ = rktlog.NewLogSet("stage1 gc", debug)
if !debug {
diag.SetOutput(ioutil.Discard)
}
podID, err := types.NewUUID(flag.Arg(0))
if err != nil {
log.Fatal("UUID is missing or malformed")
}
diag.Printf("Removing journal link.")
if err := removeJournalLink(podID); err != nil {
log.PrintE("error removing journal link", err)
}
diag.Printf("Cleaning up cgroups.")
if err := cleanupV1Cgroups(); err != nil {
log.PrintE("error cleaning up cgroups", err)
}
diag.Printf("Tearing down networks.")
if err := gcNetworking(podID); err != nil {
log.FatalE("", err)
}
}
func handleRegisterPod(w http.ResponseWriter, r *http.Request) {
defer r.Body.Close()
uuid, err := types.NewUUID(mux.Vars(r)["uuid"])
if err != nil {
w.WriteHeader(http.StatusBadRequest)
fmt.Fprintf(w, "UUID is missing or malformed: %v", err)
return
}
token := queryValue(r.URL, "token")
if token == "" {
w.WriteHeader(http.StatusBadRequest)
fmt.Fprint(w, "token missing")
return
}
pm := &schema.PodManifest{}
if err := json.NewDecoder(r.Body).Decode(pm); err != nil {
w.WriteHeader(http.StatusBadRequest)
fmt.Fprintf(w, "JSON-decoding failed: %v", err)
return
}
pods.addPod(uuid, token, pm)
w.WriteHeader(http.StatusOK)
}
func (s *v1AlphaAPIServer) InspectPod(ctx context.Context, request *v1alpha.InspectPodRequest) (*v1alpha.InspectPodResponse, error) {
uuid, err := types.NewUUID(request.Id)
if err != nil {
stderr.PrintE(fmt.Sprintf("invalid pod id %q", request.Id), err)
return nil, err
}
p, err := getPod(uuid)
if err != nil {
stderr.PrintE(fmt.Sprintf("failed to get pod %q", request.Id), err)
return nil, err
}
defer p.Close()
pod, _, err := getBasicPod(p)
if err != nil {
return nil, err
}
// Fill the extra pod info that is not available in ListPods(detail=false).
if err := fillAppInfo(s.store, p, pod); err != nil {
return nil, err
}
return &v1alpha.InspectPodResponse{Pod: pod}, nil
}
func handleRegisterApp(w http.ResponseWriter, r *http.Request) {
defer r.Body.Close()
uuid, err := types.NewUUID(mux.Vars(r)["uuid"])
if err != nil {
w.WriteHeader(http.StatusBadRequest)
fmt.Fprintf(w, "UUID is missing or malformed: %v", err)
return
}
an := mux.Vars(r)["app"]
if an == "" {
w.WriteHeader(http.StatusBadRequest)
fmt.Fprint(w, "app missing")
return
}
im := &schema.ImageManifest{}
if err := json.NewDecoder(r.Body).Decode(im); err != nil {
w.WriteHeader(http.StatusBadRequest)
fmt.Fprintf(w, "JSON-decoding failed: %v", err)
return
}
err = pods.addApp(uuid, an, im)
if err != nil {
w.WriteHeader(http.StatusNotFound)
fmt.Fprint(w, "Pod with given UUID not found")
return
}
w.WriteHeader(http.StatusOK)
}
func TestImagePrepareRmIDRun(t *testing.T) {
imageFile := patchTestACI(unreferencedACI, fmt.Sprintf("--name=%s", unreferencedApp))
defer os.Remove(imageFile)
ctx := testutils.NewRktRunCtx()
defer ctx.Cleanup()
cmd := fmt.Sprintf("%s --insecure-options=image fetch %s", ctx.Cmd(), imageFile)
t.Logf("Fetching %s", imageFile)
spawnAndWaitOrFail(t, cmd, 0)
// at this point we know that RKT_INSPECT_IMAGE env var is not empty
referencedACI := os.Getenv("RKT_INSPECT_IMAGE")
cmds := strings.Fields(ctx.Cmd())
prepareCmd := exec.Command(cmds[0], cmds[1:]...)
prepareCmd.Args = append(prepareCmd.Args, "--insecure-options=image", "prepare", referencedACI)
output, err := prepareCmd.Output()
if err != nil {
t.Fatalf("Cannot read the output: %v", err)
}
podIDStr := strings.TrimSpace(string(output))
podID, err := types.NewUUID(podIDStr)
if err != nil {
t.Fatalf("%q is not a valid UUID: %v", podIDStr, err)
}
t.Logf("Retrieving stage1 imageID")
stage1ImageID, err := getImageID(ctx, stage1App)
if err != nil {
t.Fatalf("rkt didn't terminate correctly: %v", err)
}
t.Logf("Retrieving %s image ID", referencedApp)
referencedImageID, err := getImageID(ctx, referencedApp)
if err != nil {
t.Fatalf("rkt didn't terminate correctly: %v", err)
}
t.Logf("Retrieving %s image ID", unreferencedApp)
unreferencedImageID, err := getImageID(ctx, unreferencedApp)
if err != nil {
t.Fatalf("rkt didn't terminate correctly: %v", err)
}
t.Logf("Removing stage1 image (should work)")
removeImage(t, ctx, stage1ImageID)
t.Logf("Removing image for app %s (should work)", referencedApp)
removeImage(t, ctx, referencedImageID)
t.Logf("Removing image for app %s (should work)", unreferencedApp)
removeImage(t, ctx, unreferencedImageID)
cmd = fmt.Sprintf("%s run-prepared --mds-register=false %s", ctx.Cmd(), podID.String())
t.Logf("Running %s", referencedACI)
spawnAndWaitOrFail(t, cmd, 0)
}
func readUUIDFromFile(path string) (*types.UUID, error) {
uuid, err := ioutil.ReadFile(path)
if err != nil {
return nil, err
}
uuid = bytes.TrimSpace(uuid)
return types.NewUUID(string(uuid))
}
func TestPodStoreAddApp(t *testing.T) {
ps, _, _, app := setupPodStoreTest(t)
uuid2, err := types.NewUUID("fe305d54-75b4-431b-adb2-eb6b9e546013")
if err != nil {
panic("bad uuid literal")
}
im := &schema.ImageManifest{}
if err = ps.addApp(uuid2, app, im); err != errPodNotFound {
t.Errorf("addApp with unknown pod returned: %v", err)
}
}
func main() {
flag.Parse()
log := rktlog.New(os.Stderr, "stage1 gc", debug)
podID, err := types.NewUUID(flag.Arg(0))
if err != nil {
log.Fatal("UUID is missing or malformed")
}
if err := gcNetworking(podID); err != nil {
log.FatalE("", err)
}
}
func validatePodMetadata(metadataURL string, pm *schema.PodManifest) results {
r := results{}
uuid, err := metadataGet(metadataURL, "/pod/uuid")
if err != nil {
return append(r, err)
}
_, err = types.NewUUID(string(uuid))
if err != nil {
return append(r, fmt.Errorf("malformed UUID returned (%v): %v", string(uuid), err))
}
return append(r, validatePodAnnotations(metadataURL, pm)...)
}
func runRktAndGetUUID(t *testing.T, rktCmd string) string {
child := spawnOrFail(t, rktCmd)
defer waitOrFail(t, child, 0)
result, out, err := expectRegexWithOutput(child, "\n[0-9a-f-]{36}")
if err != nil || len(result) != 1 {
t.Fatalf("Error: %v\nOutput: %v", err, out)
}
podIDStr := strings.TrimSpace(result[0])
podID, err := types.NewUUID(podIDStr)
if err != nil {
t.Fatalf("%q is not a valid UUID: %v", podIDStr, err)
}
return podID.String()
}
func handleUnregisterPod(w http.ResponseWriter, r *http.Request) {
defer r.Body.Close()
uuid, err := types.NewUUID(mux.Vars(r)["uuid"])
if err != nil {
w.WriteHeader(http.StatusBadRequest)
fmt.Fprintf(w, "UUID is missing or malformed: %v", err)
return
}
if err := pods.remove(uuid); err != nil {
w.WriteHeader(http.StatusNotFound)
fmt.Fprint(w, err)
return
}
w.WriteHeader(http.StatusOK)
}
// walkPods iterates over the included directories calling function f for every pod found.
func walkPods(include includeMask, f func(*pod)) error {
if err := initPods(); err != nil {
return err
}
ls, err := listPods(include)
if err != nil {
return errwrap.Wrap(errors.New("failed to get pods"), err)
}
sort.Strings(ls)
for _, uuid := range ls {
u, err := types.NewUUID(uuid)
if err != nil {
stderr.PrintE(fmt.Sprintf("skipping %q", uuid), err)
continue
}
p, err := getPod(u)
if err != nil {
stderr.PrintE(fmt.Sprintf("skipping %q", uuid), err)
continue
}
// omit pods found in unrequested states
// this is to cover a race between listPods finding the uuids and pod states changing
// it's preferable to keep these operations lock-free, for example a `rkt gc` shouldn't block `rkt run`.
if p.isEmbryo && include&includeEmbryoDir == 0 ||
p.isExitedGarbage && include&includeExitedGarbageDir == 0 ||
p.isGarbage && include&includeGarbageDir == 0 ||
p.isPrepared && include&includePreparedDir == 0 ||
((p.isPreparing || p.isAbortedPrepare) && include&includePrepareDir == 0) ||
p.isRunning() && include&includeRunDir == 0 {
p.Close()
continue
}
f(p)
p.Close()
}
return nil
}
func (s *v1AlphaAPIServer) InspectPod(ctx context.Context, request *v1alpha.InspectPodRequest) (*v1alpha.InspectPodResponse, error) {
uuid, err := types.NewUUID(request.Id)
if err != nil {
stderr.PrintE(fmt.Sprintf("invalid pod id %q", request.Id), err)
return nil, err
}
p, err := getPod(uuid)
if err != nil {
stderr.PrintE(fmt.Sprintf("failed to get pod %q", request.Id), err)
return nil, err
}
defer p.Close()
pod := s.getBasicPod(p)
fillPodDetails(s.store, p, pod)
return &v1alpha.InspectPodResponse{Pod: pod}, nil
}
func main() {
flag.Parse()
podID, err := types.NewUUID(flag.Arg(0))
if err != nil {
log.Fatal("UUID is missing or malformed")
}
if err := removeJournalLink(podID); err != nil {
log.PrintE("error removing journal link", err)
}
if err := cleanupV1Cgroups(); err != nil {
log.PrintE("error cleaning up cgroups", err)
}
if err := gcNetworking(podID); err != nil {
log.FatalE("", err)
}
}
func ProcessArgsAndReturnPodUUID() *types.UUID {
flag.Parse()
if cliDebugFlag {
logs.SetLevel(logs.DEBUG)
}
if lvlStr := os.Getenv(common.EnvLogLevel); lvlStr != "" {
lvl, err := logs.ParseLevel(lvlStr)
if err != nil {
fmt.Printf("Unknown log level : %s", lvlStr)
os.Exit(1)
}
logs.SetLevel(lvl)
}
arg := flag.Arg(0)
uuid, err := types.NewUUID(arg)
if err != nil {
logs.WithE(err).WithField("content", arg).Fatal("UUID is missing or malformed")
}
return uuid
}
// resolveUUID attempts to resolve the uuid specified as uuid against all pods present.
// An unambiguously matched uuid or nil is returned.
func resolveUUID(dataDir, uuid string) (*types.UUID, error) {
uuid = strings.ToLower(uuid)
m, err := matchUUID(dataDir, uuid)
if err != nil {
return nil, err
}
if len(m) == 0 {
return nil, fmt.Errorf("no matches found for %q", uuid)
}
if len(m) > 1 {
return nil, fmt.Errorf("ambiguous uuid, %d matches", len(m))
}
u, err := types.NewUUID(m[0])
if err != nil {
return nil, errwrap.Wrap(errors.New("invalid UUID"), err)
}
return u, nil
}
func setupPodStoreTest(t *testing.T) (*podStore, *types.UUID, string, string) {
ps := newPodStore()
uuid, err := types.NewUUID("de305d54-75b4-431b-adb2-eb6b9e546013")
if err != nil {
panic("bad uuid literal")
}
ip := "1.2.3.4"
app := "myapp"
pm := &schema.PodManifest{}
ps.addPod(uuid, ip, pm)
im := &schema.ImageManifest{}
err = ps.addApp(uuid, app, im)
if err != nil {
t.Fatalf("addApp failed with %v", err)
}
return ps, uuid, ip, app
}
// NewPod creates a new pod directory in the "preparing" state, allocating a unique uuid for it in the process.
// The returned pod is always left in an exclusively locked state (preparing is locked in the prepared directory)
// The pod must be closed using pod.Close()
func NewPod(dataDir string) (*Pod, error) {
if err := initPods(dataDir); err != nil {
return nil, err
}
p := &Pod{
dataDir: dataDir,
createdByMe: true,
isEmbryo: true, // starts as an embryo, then ToPreparing locks, renames, and sets isPreparing
// rest start false.
}
var err error
p.UUID, err = types.NewUUID(uuid.New())
if err != nil {
return nil, errwrap.Wrap(errors.New("error creating UUID"), err)
}
err = os.Mkdir(p.embryoPath(), 0750)
if err != nil {
return nil, err
}
p.FileLock, err = lock.NewLock(p.embryoPath(), lock.Dir)
if err != nil {
os.Remove(p.embryoPath())
return nil, err
}
err = p.ToPreparing()
if err != nil {
return nil, err
}
// At this point we we have:
// /var/lib/rkt/pods/prepare/$uuid << exclusively locked to indicate "preparing"
return p, nil
}
func main() {
flag.Parse()
stage1initcommon.InitDebug(debug)
log, diag, _ = rktlog.NewLogSet("app-add", debug)
if !debug {
diag.SetOutput(ioutil.Discard)
}
uuid, err := types.NewUUID(flagUUID)
if err != nil {
log.FatalE("UUID is missing or malformed", err)
}
appName, err := types.NewACName(flagApp)
if err != nil {
log.FatalE("invalid app name", err)
}
root := "."
p, err := stage1types.LoadPod(root, uuid)
if err != nil {
log.FatalE("failed to load pod", err)
}
flavor, _, err := stage1initcommon.GetFlavor(p)
if err != nil {
log.FatalE("failed to get stage1 flavor", err)
}
insecureOptions := stage1initcommon.Stage1InsecureOptions{
DisablePaths: disablePaths,
DisableCapabilities: disableCapabilities,
DisableSeccomp: disableSeccomp,
}
ra := p.Manifest.Apps.Get(*appName)
if ra == nil {
log.Fatalf("failed to find app %q", *appName)
}
binPath, err := stage1initcommon.FindBinPath(p, ra)
if err != nil {
log.FatalE("failed to find bin path", err)
}
if ra.App.WorkingDirectory == "" {
ra.App.WorkingDirectory = "/"
}
enterCmd := stage1common.PrepareEnterCmd(false)
stage1initcommon.AppAddMounts(p, ra, enterCmd)
// when using host cgroups, make the subgroup writable by pod systemd
if flavor != "kvm" {
err = prepareAppCgroups(p, ra, enterCmd)
if err != nil {
log.FatalE("error preparing cgroups", err)
}
}
// write service files
w := stage1initcommon.NewUnitWriter(p)
w.AppUnit(ra, binPath, privateUsers, insecureOptions,
unit.NewUnitOption("Unit", "Before", "halt.target"),
unit.NewUnitOption("Unit", "Conflicts", "halt.target"),
unit.NewUnitOption("Service", "StandardOutput", "journal+console"),
unit.NewUnitOption("Service", "StandardError", "journal+console"),
)
w.AppReaperUnit(ra.Name, binPath)
if err := w.Error(); err != nil {
log.FatalE("error generating app units", err)
}
// stage2 environment is ready at this point, but systemd does not know
// about the new application yet
args := enterCmd
args = append(args, "/usr/bin/systemctl")
args = append(args, "daemon-reload")
cmd := exec.Cmd{
Path: args[0],
Args: args,
}
if out, err := cmd.CombinedOutput(); err != nil {
log.Fatalf("%q failed at daemon-reload:\n%s", appName, out)
}
os.Exit(0)
}
func stage1(rp *stage1commontypes.RuntimePod) int {
uuid, err := types.NewUUID(flag.Arg(0))
if err != nil {
log.Print("UUID is missing or malformed\n")
return 254
}
root := "."
p, err := stage1commontypes.LoadPod(root, uuid, rp)
if err != nil {
log.PrintE("can't load pod", err)
return 254
}
if err := p.SaveRuntime(); err != nil {
log.FatalE("failed to save runtime parameters", err)
}
// Sanity checks
if len(p.Manifest.Apps) != 1 {
log.Printf("flavor %q only supports 1 application per Pod for now", flavor)
return 254
}
ra := p.Manifest.Apps[0]
imgName := p.AppNameToImageName(ra.Name)
args := ra.App.Exec
if len(args) == 0 {
log.Printf(`image %q has an empty "exec" (try --exec=BINARY)`, imgName)
return 254
}
lfd, err := common.GetRktLockFD()
if err != nil {
log.PrintE("can't get rkt lock fd", err)
return 254
}
// set close-on-exec flag on RKT_LOCK_FD so it gets correctly closed after execution is finished
if err := sys.CloseOnExec(lfd, true); err != nil {
log.PrintE("can't set FD_CLOEXEC on rkt lock", err)
return 254
}
workDir := "/"
if ra.App.WorkingDirectory != "" {
workDir = ra.App.WorkingDirectory
}
env := []string{"PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin"}
for _, e := range ra.App.Environment {
env = append(env, e.Name+"="+e.Value)
}
rfs := filepath.Join(common.AppPath(p.Root, ra.Name), "rootfs")
argFlyMounts, err := evaluateMounts(rfs, string(ra.Name), p)
if err != nil {
log.PrintE("can't evaluate mounts", err)
return 254
}
effectiveMounts := append(
[]flyMount{
{"", "", "/dev", "none", syscall.MS_REC | syscall.MS_SHARED},
{"/dev", rfs, "/dev", "none", syscall.MS_BIND | syscall.MS_REC},
{"", "", "/proc", "none", syscall.MS_REC | syscall.MS_SHARED},
{"/proc", rfs, "/proc", "none", syscall.MS_BIND | syscall.MS_REC},
{"", "", "/sys", "none", syscall.MS_REC | syscall.MS_SHARED},
{"/sys", rfs, "/sys", "none", syscall.MS_BIND | syscall.MS_REC},
{"tmpfs", rfs, "/tmp", "tmpfs", 0},
},
argFlyMounts...,
)
/* Process DNS config files
*
* /etc/resolv.conf: four modes
* 'host' - bind-mount host's file
* 'stage0' - bind-mount the file created by stage0
* 'default' - do nothing (we would respect CNI if fly had networking)
* 'none' - do nothing
*/
switch p.ResolvConfMode {
case "host":
effectiveMounts = append(effectiveMounts,
flyMount{"/etc/resolv.conf", rfs, "/etc/resolv.conf", "none", syscall.MS_BIND | syscall.MS_RDONLY})
case "stage0":
if err := copyResolv(p); err != nil {
log.PrintE("can't copy /etc/resolv.conf", err)
return 254
}
}
/*
* /etc/hosts: three modes:
* 'host' - bind-mount hosts's file
* 'stage0' - bind mount the file created by stage1
* 'default' - create a stub /etc/hosts if needed
*/
switch p.EtcHostsMode {
case "host":
effectiveMounts = append(effectiveMounts,
flyMount{"/etc/hosts", rfs, "/etc/hosts", "none", syscall.MS_BIND | syscall.MS_RDONLY})
case "stage0":
effectiveMounts = append(effectiveMounts, flyMount{
filepath.Join(common.Stage1RootfsPath(p.Root), "etc", "rkt-hosts"),
rfs,
"/etc/hosts",
"none",
syscall.MS_BIND | syscall.MS_RDONLY})
case "default":
stage2HostsPath := filepath.Join(common.AppRootfsPath(p.Root, ra.Name), "etc", "hosts")
if _, err := os.Stat(stage2HostsPath); err != nil && os.IsNotExist(err) {
fallbackHosts := []byte("127.0.0.1 localhost localdomain\n")
ioutil.WriteFile(stage2HostsPath, fallbackHosts, 0644)
}
}
for _, mount := range effectiveMounts {
diag.Printf("Processing %+v", mount)
var (
err error
hostPathInfo os.FileInfo
targetPathInfo os.FileInfo
)
if strings.HasPrefix(mount.HostPath, "/") {
if hostPathInfo, err = os.Stat(mount.HostPath); err != nil {
log.PrintE(fmt.Sprintf("stat of host path %s", mount.HostPath), err)
return 254
}
} else {
hostPathInfo = nil
}
absTargetPath := filepath.Join(mount.TargetPrefixPath, mount.RelTargetPath)
if targetPathInfo, err = os.Stat(absTargetPath); err != nil && !os.IsNotExist(err) {
log.PrintE(fmt.Sprintf("stat of target path %s", absTargetPath), err)
return 254
}
switch {
case (mount.Flags & syscall.MS_REMOUNT) != 0:
{
diag.Printf("don't attempt to create files for remount of %q", absTargetPath)
}
case targetPathInfo == nil:
absTargetPathParent, _ := filepath.Split(absTargetPath)
if err := os.MkdirAll(absTargetPathParent, 0755); err != nil {
log.PrintE(fmt.Sprintf("can't create directory %q", absTargetPath), err)
return 254
}
switch {
case hostPathInfo == nil || hostPathInfo.IsDir():
if err := os.Mkdir(absTargetPath, 0755); err != nil {
log.PrintE(fmt.Sprintf("can't create directory %q", absTargetPath), err)
return 254
}
case !hostPathInfo.IsDir():
file, err := os.OpenFile(absTargetPath, os.O_CREATE, 0700)
if err != nil {
log.PrintE(fmt.Sprintf("can't create file %q", absTargetPath), err)
return 254
}
file.Close()
}
case hostPathInfo != nil:
switch {
case hostPathInfo.IsDir() && !targetPathInfo.IsDir():
log.Printf("can't mount because %q is a directory while %q is not", mount.HostPath, absTargetPath)
return 254
case !hostPathInfo.IsDir() && targetPathInfo.IsDir():
log.Printf("can't mount because %q is not a directory while %q is", mount.HostPath, absTargetPath)
return 254
}
}
if err := syscall.Mount(mount.HostPath, absTargetPath, mount.Fs, mount.Flags, ""); err != nil {
log.PrintE(fmt.Sprintf("can't mount %q on %q with flags %v", mount.HostPath, absTargetPath, mount.Flags), err)
return 254
}
}
if err = stage1common.WritePid(os.Getpid(), "pid"); err != nil {
log.Error(err)
return 254
}
var uidResolver, gidResolver user.Resolver
var uid, gid int
uidResolver, err = user.NumericIDs(ra.App.User)
if err != nil {
uidResolver, err = user.IDsFromStat(rfs, ra.App.User, nil)
}
if err != nil { // give up
log.PrintE(fmt.Sprintf("invalid user %q", ra.App.User), err)
return 254
}
if uid, _, err = uidResolver.IDs(); err != nil {
log.PrintE(fmt.Sprintf("failed to configure user %q", ra.App.User), err)
return 254
}
gidResolver, err = user.NumericIDs(ra.App.Group)
if err != nil {
gidResolver, err = user.IDsFromStat(rfs, ra.App.Group, nil)
}
if err != nil { // give up
log.PrintE(fmt.Sprintf("invalid group %q", ra.App.Group), err)
return 254
}
if _, gid, err = gidResolver.IDs(); err != nil {
log.PrintE(fmt.Sprintf("failed to configure group %q", ra.App.Group), err)
return 254
}
diag.Printf("chroot to %q", rfs)
if err := syscall.Chroot(rfs); err != nil {
log.PrintE("can't chroot", err)
return 254
}
if err := os.Chdir(workDir); err != nil {
log.PrintE(fmt.Sprintf("can't change to working directory %q", workDir), err)
return 254
}
// lock the current goroutine to its current OS thread.
// This will force the subsequent syscalls to be executed in the same OS thread as Setresuid, and Setresgid,
// see https://github.com/golang/go/issues/1435#issuecomment-66054163.
runtime.LockOSThread()
diag.Printf("setting uid %d gid %d", uid, gid)
if err := syscall.Setresgid(gid, gid, gid); err != nil {
log.PrintE(fmt.Sprintf("can't set gid %d", gid), err)
return 254
}
if err := syscall.Setresuid(uid, uid, uid); err != nil {
log.PrintE(fmt.Sprintf("can't set uid %d", uid), err)
return 254
}
diag.Printf("execing %q in %q", args, rfs)
err = stage1common.WithClearedCloExec(lfd, func() error {
return syscall.Exec(args[0], args, env)
})
if err != nil {
log.PrintE(fmt.Sprintf("can't execute %q", args[0]), err)
return 254
}
return 0
}
func (s *v1AlphaAPIServer) constrainedGetLogs(request *v1alpha.GetLogsRequest, server v1alpha.PublicAPI_GetLogsServer) error {
uuid, err := types.NewUUID(request.PodId)
if err != nil {
return err
}
pod, err := getPod(uuid)
if err != nil {
return err
}
defer pod.Close()
stage1Path := "stage1/rootfs"
if pod.usesOverlay() {
stage1TreeStoreID, err := pod.getStage1TreeStoreID()
if err != nil {
return err
}
stage1Path = fmt.Sprintf("/overlay/%s/upper/", stage1TreeStoreID)
}
path := filepath.Join(getDataDir(), "/pods/run/", request.PodId, stage1Path, "/var/log/journal/")
if _, err := os.Stat(path); os.IsNotExist(err) {
return fmt.Errorf("%s: logging unsupported", uuid.String())
}
jconf := sdjournal.JournalReaderConfig{
Path: path,
}
if request.AppName != "" {
jconf.Matches = []sdjournal.Match{
{
Field: sdjournal.SD_JOURNAL_FIELD_SYSLOG_IDENTIFIER,
Value: request.AppName,
},
}
}
if request.SinceTime != 0 {
t := time.Unix(request.SinceTime, 0)
jconf.Since = -time.Since(t)
}
if request.Lines != 0 {
jconf.NumFromTail = uint64(request.Lines)
}
jr, err := sdjournal.NewJournalReader(jconf)
if err != nil {
return err
}
defer jr.Close()
if request.Follow {
return jr.Follow(nil, LogsStreamWriter{server: server})
}
data, err := ioutil.ReadAll(jr)
if err != nil {
return err
}
return server.Send(&v1alpha.GetLogsResponse{Lines: common.RemoveEmptyLines(string(data))})
}
func TestGetPodAndRefreshState(t *testing.T) {
testCases := []struct {
paths []dirFn
locks []dirFn
expected states
}{
{
paths: []dirFn{embryoDir},
expected: states{isEmbryo: true},
},
{
paths: []dirFn{prepareDir},
locks: []dirFn{prepareDir},
expected: states{isPreparing: true},
},
{
paths: []dirFn{prepareDir},
expected: states{isAbortedPrepare: true},
},
{
paths: []dirFn{runDir},
locks: []dirFn{runDir},
expected: states{},
},
{
paths: []dirFn{runDir},
expected: states{isExited: true},
},
{
paths: []dirFn{garbageDir},
expected: states{isGarbage: true},
},
{
paths: []dirFn{garbageDir},
locks: []dirFn{garbageDir},
expected: states{isGarbage: true, isDeleting: true},
},
}
uuid, err := types.NewUUID("aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaaa")
if err != nil {
panic(err)
}
for i, tcase := range testCases {
tmpDir, err := ioutil.TempDir("", "")
if err != nil {
panic(err)
}
defer os.RemoveAll(tmpDir)
for _, pfn := range tcase.paths {
podPath := filepath.Join(pfn(tmpDir), uuid.String())
if err := os.MkdirAll(podPath, 0777); err != nil {
panic(err)
}
}
for _, lfn := range tcase.locks {
podPath := filepath.Join(lfn(tmpDir), uuid.String())
l, err := lock.NewLock(podPath, lock.Dir)
if err != nil {
t.Fatalf("error taking lock on directory: %v", err)
}
err = l.ExclusiveLock()
if err != nil {
t.Fatalf("could not get exclusive lock on directory: %v", err)
}
defer l.Unlock()
}
p, err := getPod(tmpDir, uuid)
if err != nil {
t.Fatalf("%v: unable to get pod: %v", i, err)
}
pstate := podToStates(p)
if !reflect.DeepEqual(tcase.expected, pstate) {
t.Errorf("%v: expected %+v == %+v after getPod", i, tcase.expected, pstate)
}
err = p.refreshState()
if err != nil {
t.Errorf("error refreshing state: %v", err)
continue
}
pstate = podToStates(p)
if !reflect.DeepEqual(tcase.expected, pstate) {
t.Errorf("%v: expected %+v == %+v after refrshState", i, tcase.expected, pstate)
}
}
}
func stage1() int {
uuid, err := types.NewUUID(flag.Arg(0))
if err != nil {
log.PrintE("UUID is missing or malformed", err)
return 1
}
root := "."
p, err := stage1commontypes.LoadPod(root, uuid)
if err != nil {
log.PrintE("failed to load pod", err)
return 1
}
// set close-on-exec flag on RKT_LOCK_FD so it gets correctly closed when invoking
// network plugins
lfd, err := common.GetRktLockFD()
if err != nil {
log.PrintE("failed to get rkt lock fd", err)
return 1
}
if err := sys.CloseOnExec(lfd, true); err != nil {
log.PrintE("failed to set FD_CLOEXEC on rkt lock", err)
return 1
}
mirrorLocalZoneInfo(p.Root)
flavor, _, err := stage1initcommon.GetFlavor(p)
if err != nil {
log.PrintE("failed to get stage1 flavor", err)
return 3
}
var n *networking.Networking
if netList.Contained() {
fps, err := forwardedPorts(p)
if err != nil {
log.Error(err)
return 6
}
n, err = networking.Setup(root, p.UUID, fps, netList, localConfig, flavor, debug)
if err != nil {
log.PrintE("failed to setup network", err)
return 6
}
if err = n.Save(); err != nil {
log.PrintE("failed to save networking state", err)
n.Teardown(flavor, debug)
return 6
}
if len(mdsToken) > 0 {
hostIP, err := n.GetDefaultHostIP()
if err != nil {
log.PrintE("failed to get default Host IP", err)
return 6
}
p.MetadataServiceURL = common.MetadataServicePublicURL(hostIP, mdsToken)
}
} else {
if flavor == "kvm" {
log.Print("flavor kvm requires private network configuration (try --net)")
return 6
}
if len(mdsToken) > 0 {
p.MetadataServiceURL = common.MetadataServicePublicURL(localhostIP, mdsToken)
}
}
if err = stage1initcommon.WriteDefaultTarget(p); err != nil {
log.PrintE("failed to write default.target", err)
return 2
}
if err = stage1initcommon.WritePrepareAppTemplate(p); err != nil {
log.PrintE("failed to write prepare-app service template", err)
return 2
}
if err := stage1initcommon.SetJournalPermissions(p); err != nil {
log.PrintE("warning: error setting journal ACLs, you'll need root to read the pod journal", err)
}
if flavor == "kvm" {
if err := KvmPodToSystemd(p, n); err != nil {
log.PrintE("failed to configure systemd for kvm", err)
return 2
}
}
if err = stage1initcommon.PodToSystemd(p, interactive, flavor, privateUsers); err != nil {
log.PrintE("failed to configure systemd", err)
return 2
}
args, env, err := getArgsEnv(p, flavor, debug, n)
if err != nil {
log.Error(err)
return 3
}
// create a separate mount namespace so the cgroup filesystems
// are unmounted when exiting the pod
if err := syscall.Unshare(syscall.CLONE_NEWNS); err != nil {
log.FatalE("error unsharing", err)
}
// we recursively make / a "shared and slave" so mount events from the
// new namespace don't propagate to the host namespace but mount events
// from the host propagate to the new namespace and are forwarded to
// its peer group
// See https://www.kernel.org/doc/Documentation/filesystems/sharedsubtree.txt
if err := syscall.Mount("", "/", "none", syscall.MS_REC|syscall.MS_SLAVE, ""); err != nil {
log.FatalE("error making / a slave mount", err)
}
if err := syscall.Mount("", "/", "none", syscall.MS_REC|syscall.MS_SHARED, ""); err != nil {
log.FatalE("error making / a shared and slave mount", err)
}
enabledCgroups, err := cgroup.GetEnabledCgroups()
if err != nil {
log.FatalE("error getting cgroups", err)
return 5
}
// mount host cgroups in the rkt mount namespace
if err := mountHostCgroups(enabledCgroups); err != nil {
log.FatalE("couldn't mount the host cgroups", err)
return 5
}
var serviceNames []string
for _, app := range p.Manifest.Apps {
serviceNames = append(serviceNames, stage1initcommon.ServiceUnitName(app.Name))
}
s1Root := common.Stage1RootfsPath(p.Root)
machineID := stage1initcommon.GetMachineID(p)
subcgroup, err := getContainerSubCgroup(machineID)
if err == nil {
if err := mountContainerCgroups(s1Root, enabledCgroups, subcgroup, serviceNames); err != nil {
log.PrintE("couldn't mount the container cgroups", err)
return 5
}
} else {
log.PrintE("continuing with per-app isolators disabled", err)
}
if err = stage1common.WritePpid(os.Getpid()); err != nil {
log.Error(err)
return 4
}
err = stage1common.WithClearedCloExec(lfd, func() error {
return syscall.Exec(args[0], args, env)
})
if err != nil {
log.PrintE(fmt.Sprintf("failed to execute %q", args[0]), err)
return 7
}
return 0
}
func stage1() int {
uuid, err := types.NewUUID(flag.Arg(0))
if err != nil {
log.FatalE("UUID is missing or malformed", err)
}
root := "."
p, err := stage1commontypes.LoadPod(root, uuid)
if err != nil {
log.FatalE("failed to load pod", err)
}
// set close-on-exec flag on RKT_LOCK_FD so it gets correctly closed when invoking
// network plugins
lfd, err := common.GetRktLockFD()
if err != nil {
log.FatalE("failed to get rkt lock fd", err)
}
if err := sys.CloseOnExec(lfd, true); err != nil {
log.FatalE("failed to set FD_CLOEXEC on rkt lock", err)
}
mirrorLocalZoneInfo(p.Root)
flavor, _, err := stage1initcommon.GetFlavor(p)
if err != nil {
log.FatalE("failed to get stage1 flavor", err)
}
var n *networking.Networking
if netList.Contained() {
fps, err := commonnet.ForwardedPorts(p.Manifest)
if err != nil {
log.FatalE("error initializing forwarding ports", err)
}
noDNS := dnsConfMode.Pairs["resolv"] != "default" // force ignore CNI DNS results
n, err = networking.Setup(root, p.UUID, fps, netList, localConfig, flavor, noDNS, debug)
if err != nil {
log.FatalE("failed to setup network", err)
}
if err = n.Save(); err != nil {
log.PrintE("failed to save networking state", err)
n.Teardown(flavor, debug)
return 254
}
if len(mdsToken) > 0 {
hostIP, err := n.GetForwardableNetHostIP()
if err != nil {
log.FatalE("failed to get default Host IP", err)
}
p.MetadataServiceURL = common.MetadataServicePublicURL(hostIP, mdsToken)
}
} else {
if flavor == "kvm" {
log.Fatal("flavor kvm requires private network configuration (try --net)")
}
if len(mdsToken) > 0 {
p.MetadataServiceURL = common.MetadataServicePublicURL(localhostIP, mdsToken)
}
}
insecureOptions := stage1initcommon.Stage1InsecureOptions{
DisablePaths: disablePaths,
DisableCapabilities: disableCapabilities,
DisableSeccomp: disableSeccomp,
}
mnt := fs.NewLoggingMounter(
fs.MounterFunc(syscall.Mount),
fs.UnmounterFunc(syscall.Unmount),
diag.Printf,
)
if dnsConfMode.Pairs["resolv"] == "host" {
stage1initcommon.UseHostResolv(mnt, root)
}
if dnsConfMode.Pairs["hosts"] == "host" {
stage1initcommon.UseHostHosts(mnt, root)
}
if mutable {
if err = stage1initcommon.MutableEnv(p); err != nil {
log.FatalE("cannot initialize mutable environment", err)
}
} else {
if err = stage1initcommon.ImmutableEnv(p, interactive, privateUsers, insecureOptions); err != nil {
log.FatalE("cannot initialize immutable environment", err)
}
}
if err := stage1initcommon.SetJournalPermissions(p); err != nil {
log.PrintE("warning: error setting journal ACLs, you'll need root to read the pod journal", err)
}
if flavor == "kvm" {
kvm.InitDebug(debug)
if err := KvmNetworkingToSystemd(p, n); err != nil {
log.FatalE("failed to configure systemd for kvm", err)
}
}
canMachinedRegister := false
if flavor != "kvm" {
// kvm doesn't register with systemd right now, see #2664.
canMachinedRegister = machinedRegister()
}
diag.Printf("canMachinedRegister %t", canMachinedRegister)
args, env, err := getArgsEnv(p, flavor, canMachinedRegister, debug, n, insecureOptions)
if err != nil {
log.FatalE("cannot get environment", err)
}
diag.Printf("args %q", args)
diag.Printf("env %q", env)
// create a separate mount namespace so the cgroup filesystems
// are unmounted when exiting the pod
if err := syscall.Unshare(syscall.CLONE_NEWNS); err != nil {
log.FatalE("error unsharing", err)
}
// we recursively make / a "shared and slave" so mount events from the
// new namespace don't propagate to the host namespace but mount events
// from the host propagate to the new namespace and are forwarded to
// its peer group
// See https://www.kernel.org/doc/Documentation/filesystems/sharedsubtree.txt
if err := mnt.Mount("", "/", "none", syscall.MS_REC|syscall.MS_SLAVE, ""); err != nil {
log.FatalE("error making / a slave mount", err)
}
if err := mnt.Mount("", "/", "none", syscall.MS_REC|syscall.MS_SHARED, ""); err != nil {
log.FatalE("error making / a shared and slave mount", err)
}
unifiedCgroup, err := cgroup.IsCgroupUnified("/")
if err != nil {
log.FatalE("error determining cgroup version", err)
}
diag.Printf("unifiedCgroup %t", unifiedCgroup)
s1Root := common.Stage1RootfsPath(p.Root)
machineID := stage1initcommon.GetMachineID(p)
subcgroup, err := getContainerSubCgroup(machineID, canMachinedRegister, unifiedCgroup)
if err != nil {
log.FatalE("error getting container subcgroup", err)
}
diag.Printf("subcgroup %q", subcgroup)
if err := ioutil.WriteFile(filepath.Join(p.Root, "subcgroup"),
[]byte(fmt.Sprintf("%s", subcgroup)), 0644); err != nil {
log.FatalE("cannot write subcgroup file", err)
}
if !unifiedCgroup {
enabledCgroups, err := v1.GetEnabledCgroups()
if err != nil {
log.FatalE("error getting v1 cgroups", err)
}
diag.Printf("enabledCgroups %q", enabledCgroups)
if err := mountHostV1Cgroups(mnt, enabledCgroups); err != nil {
log.FatalE("couldn't mount the host v1 cgroups", err)
}
if !canMachinedRegister {
if err := v1.JoinSubcgroup("systemd", subcgroup); err != nil {
log.FatalE(fmt.Sprintf("error joining subcgroup %q", subcgroup), err)
}
}
var serviceNames []string
for _, app := range p.Manifest.Apps {
serviceNames = append(serviceNames, stage1initcommon.ServiceUnitName(app.Name))
}
diag.Printf("serviceNames %q", serviceNames)
if err := mountContainerV1Cgroups(mnt, s1Root, enabledCgroups, subcgroup, serviceNames, insecureOptions); err != nil {
log.FatalE("couldn't mount the container v1 cgroups", err)
}
}
// KVM flavor has a bit different logic in handling pid vs ppid, for details look into #2389
// it doesn't require the existence of a "ppid", instead it registers the current pid (which
// will be reused by lkvm binary) as a pod process pid used during entering
pid_filename := "ppid"
if flavor == "kvm" {
pid_filename = "pid"
}
if err = stage1common.WritePid(os.Getpid(), pid_filename); err != nil {
log.FatalE("error writing pid", err)
}
if flavor == "kvm" {
if err := KvmPrepareMounts(s1Root, p); err != nil {
log.FatalE("error preparing mounts", err)
}
}
err = stage1common.WithClearedCloExec(lfd, func() error {
return syscall.Exec(args[0], args, env)
})
if err != nil {
log.FatalE(fmt.Sprintf("failed to execute %q", args[0]), err)
}
return 0
}
func stage1() int {
uuid, err := types.NewUUID(flag.Arg(0))
if err != nil {
log.Print("UUID is missing or malformed\n")
return 1
}
root := "."
p, err := stage1commontypes.LoadPod(root, uuid)
if err != nil {
log.PrintE("can't load pod", err)
return 1
}
// Sanity checks
if len(p.Manifest.Apps) != 1 {
log.Printf("flavor %q only supports 1 application per Pod for now", flavor)
return 1
}
ra := p.Manifest.Apps[0]
imgName := p.AppNameToImageName(ra.Name)
args := ra.App.Exec
if len(args) == 0 {
log.Printf(`image %q has an empty "exec" (try --exec=BINARY)`, imgName)
return 1
}
lfd, err := common.GetRktLockFD()
if err != nil {
log.PrintE("can't get rkt lock fd", err)
return 1
}
// set close-on-exec flag on RKT_LOCK_FD so it gets correctly closed after execution is finished
if err := sys.CloseOnExec(lfd, true); err != nil {
log.PrintE("can't set FD_CLOEXEC on rkt lock", err)
return 1
}
workDir := "/"
if ra.App.WorkingDirectory != "" {
workDir = ra.App.WorkingDirectory
}
env := []string{"PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin"}
for _, e := range ra.App.Environment {
env = append(env, e.Name+"="+e.Value)
}
rfs := filepath.Join(common.AppPath(p.Root, ra.Name), "rootfs")
if err := copyResolv(p); err != nil {
log.PrintE("can't copy /etc/resolv.conf", err)
return 1
}
argFlyMounts, err := evaluateMounts(rfs, string(ra.Name), p)
if err != nil {
log.PrintE("can't evaluate mounts", err)
return 1
}
effectiveMounts := append(
[]flyMount{
{"", "", "/dev", "none", syscall.MS_REC | syscall.MS_SHARED},
{"/dev", rfs, "/dev", "none", syscall.MS_BIND | syscall.MS_REC},
{"", "", "/proc", "none", syscall.MS_REC | syscall.MS_SHARED},
{"/proc", rfs, "/proc", "none", syscall.MS_BIND | syscall.MS_REC},
{"", "", "/sys", "none", syscall.MS_REC | syscall.MS_SHARED},
{"/sys", rfs, "/sys", "none", syscall.MS_BIND | syscall.MS_REC},
{"tmpfs", rfs, "/tmp", "tmpfs", 0},
},
argFlyMounts...,
)
for _, mount := range effectiveMounts {
var (
err error
hostPathInfo os.FileInfo
targetPathInfo os.FileInfo
)
if strings.HasPrefix(mount.HostPath, "/") {
if hostPathInfo, err = os.Stat(mount.HostPath); err != nil {
log.PrintE(fmt.Sprintf("stat of host path %s", mount.HostPath), err)
return 1
}
} else {
hostPathInfo = nil
}
absTargetPath := filepath.Join(mount.TargetPrefixPath, mount.RelTargetPath)
if targetPathInfo, err = os.Stat(absTargetPath); err != nil && !os.IsNotExist(err) {
log.PrintE(fmt.Sprintf("stat of target path %s", absTargetPath), err)
return 1
}
switch {
case targetPathInfo == nil:
absTargetPathParent, _ := filepath.Split(absTargetPath)
if err := os.MkdirAll(absTargetPathParent, 0755); err != nil {
log.PrintE(fmt.Sprintf("can't create directory %q", absTargetPath), err)
return 1
}
switch {
case hostPathInfo == nil || hostPathInfo.IsDir():
if err := os.Mkdir(absTargetPath, 0755); err != nil {
log.PrintE(fmt.Sprintf("can't create directory %q", absTargetPath), err)
return 1
}
case !hostPathInfo.IsDir():
file, err := os.OpenFile(absTargetPath, os.O_CREATE, 0700)
if err != nil {
log.PrintE(fmt.Sprintf("can't create file %q", absTargetPath), err)
return 1
}
file.Close()
}
case hostPathInfo != nil:
switch {
case hostPathInfo.IsDir() && !targetPathInfo.IsDir():
log.Printf("can't mount because %q is a directory while %q is not", mount.HostPath, absTargetPath)
return 1
case !hostPathInfo.IsDir() && targetPathInfo.IsDir():
log.Printf("can't mount because %q is not a directory while %q is", mount.HostPath, absTargetPath)
return 1
}
}
if err := syscall.Mount(mount.HostPath, absTargetPath, mount.Fs, mount.Flags, ""); err != nil {
log.PrintE(fmt.Sprintf("can't mount %q on %q with flags %v", mount.HostPath, absTargetPath, mount.Flags), err)
return 1
}
}
if err = stage1common.WritePid(os.Getpid(), "pid"); err != nil {
log.Error(err)
return 1
}
var uidResolver, gidResolver user.Resolver
var uid, gid int
uidResolver, err = user.NumericIDs(ra.App.User)
if err != nil {
uidResolver, err = user.IDsFromStat(rfs, ra.App.User, nil)
}
if err != nil { // give up
log.PrintE(fmt.Sprintf("invalid user %q", ra.App.User), err)
return 1
}
if uid, _, err = uidResolver.IDs(); err != nil {
log.PrintE(fmt.Sprintf("failed to configure user %q", ra.App.User), err)
return 1
}
gidResolver, err = user.NumericIDs(ra.App.Group)
if err != nil {
gidResolver, err = user.IDsFromStat(rfs, ra.App.Group, nil)
}
if err != nil { // give up
log.PrintE(fmt.Sprintf("invalid group %q", ra.App.Group), err)
return 1
}
if _, gid, err = gidResolver.IDs(); err != nil {
log.PrintE(fmt.Sprintf("failed to configure group %q", ra.App.Group), err)
return 1
}
diag.Printf("chroot to %q", rfs)
if err := syscall.Chroot(rfs); err != nil {
log.PrintE("can't chroot", err)
return 1
}
if err := os.Chdir(workDir); err != nil {
log.PrintE(fmt.Sprintf("can't change to working directory %q", workDir), err)
return 1
}
// lock the current goroutine to its current OS thread.
// This will force the subsequent syscalls to be executed in the same OS thread as Setresuid, and Setresgid,
// see https://github.com/golang/go/issues/1435#issuecomment-66054163.
runtime.LockOSThread()
diag.Printf("setting uid %d gid %d", uid, gid)
if err := syscall.Setresgid(gid, gid, gid); err != nil {
log.PrintE(fmt.Sprintf("can't set gid %d", gid), err)
return 1
}
if err := syscall.Setresuid(uid, uid, uid); err != nil {
log.PrintE(fmt.Sprintf("can't set uid %d", uid), err)
return 1
}
diag.Printf("execing %q in %q", args, rfs)
err = stage1common.WithClearedCloExec(lfd, func() error {
return syscall.Exec(args[0], args, env)
})
if err != nil {
log.PrintE(fmt.Sprintf("can't execute %q", args[0]), err)
return 1
}
return 0
}
// TODO use named flags instead of positional
func main() {
flag.Parse()
stage1initcommon.InitDebug(debug)
log, diag, _ = rktlog.NewLogSet("stage1", debug)
if !debug {
diag.SetOutput(ioutil.Discard)
}
uuid, err := types.NewUUID(flag.Arg(0))
if err != nil {
log.PrintE("UUID is missing or malformed", err)
os.Exit(254)
}
appName, err := types.NewACName(flag.Arg(1))
if err != nil {
log.PrintE("invalid app name", err)
os.Exit(254)
}
enterEP := flag.Arg(2)
root := "."
p, err := stage1types.LoadPod(root, uuid)
if err != nil {
log.PrintE("failed to load pod", err)
os.Exit(254)
}
insecureOptions := stage1initcommon.Stage1InsecureOptions{
DisablePaths: disablePaths,
DisableCapabilities: disableCapabilities,
DisableSeccomp: disableSeccomp,
}
ra := p.Manifest.Apps.Get(*appName)
if ra == nil {
log.Printf("failed to get app")
os.Exit(254)
}
if ra.App.WorkingDirectory == "" {
ra.App.WorkingDirectory = "/"
}
binPath, err := stage1initcommon.FindBinPath(p, ra)
if err != nil {
log.PrintE("failed to find bin path", err)
os.Exit(254)
}
w := stage1initcommon.NewUnitWriter(p)
w.AppUnit(ra, binPath, privateUsers, insecureOptions,
unit.NewUnitOption("Unit", "Before", "halt.target"),
unit.NewUnitOption("Unit", "Conflicts", "halt.target"),
unit.NewUnitOption("Service", "StandardOutput", "journal+console"),
unit.NewUnitOption("Service", "StandardError", "journal+console"),
)
w.AppReaperUnit(ra.Name, binPath)
if err := w.Error(); err != nil {
log.PrintE("error generating app units", err)
os.Exit(254)
}
args := []string{enterEP}
args = append(args, fmt.Sprintf("--pid=%s", flag.Arg(3)))
args = append(args, "/usr/bin/systemctl")
args = append(args, "daemon-reload")
cmd := exec.Cmd{
Path: args[0],
Args: args,
}
if err := cmd.Run(); err != nil {
log.PrintE("error executing daemon-reload", err)
os.Exit(254)
}
args = []string{enterEP}
args = append(args, fmt.Sprintf("--pid=%s", flag.Arg(3)))
args = append(args, "/usr/bin/systemctl")
args = append(args, "start")
args = append(args, appName.String())
cmd = exec.Cmd{
Path: args[0],
Args: args,
}
if err := cmd.Run(); err != nil {
log.PrintE(fmt.Sprintf("error starting app %q", appName.String()), err)
os.Exit(254)
}
// TODO unmount all the volumes
os.Exit(0)
}
// isUUID returns true if the input is a valid rkt UUID,
// e.g. "2372bc17-47cb-43fb-8d78-20b31729feda".
func isUUID(input string) bool {
if _, err := appctypes.NewUUID(input); err != nil {
return false
}
return true
}
func stage1() int {
uuid, err := types.NewUUID(flag.Arg(0))
if err != nil {
log.Print("UUID is missing or malformed\n")
return 1
}
root := "."
p, err := stage1commontypes.LoadPod(root, uuid)
if err != nil {
log.PrintE("can't load pod", err)
return 1
}
if len(p.Manifest.Apps) != 1 {
log.Printf("flavor %q only supports 1 application per Pod for now", flavor)
return 1
}
lfd, err := common.GetRktLockFD()
if err != nil {
log.PrintE("can't get rkt lock fd", err)
return 1
}
// set close-on-exec flag on RKT_LOCK_FD so it gets correctly closed after execution is finished
if err := sys.CloseOnExec(lfd, true); err != nil {
log.PrintE("can't set FD_CLOEXEC on rkt lock", err)
return 1
}
env := []string{"PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin"}
for _, e := range p.Manifest.Apps[0].App.Environment {
env = append(env, e.Name+"="+e.Value)
}
args := p.Manifest.Apps[0].App.Exec
rfs := filepath.Join(common.AppPath(p.Root, p.Manifest.Apps[0].Name), "rootfs")
argFlyMounts, err := evaluateMounts(rfs, string(p.Manifest.Apps[0].Name), p)
if err != nil {
log.PrintE("can't evaluate mounts", err)
return 1
}
effectiveMounts := append(
[]flyMount{
{"", "", "/dev", "none", syscall.MS_REC | syscall.MS_SHARED},
{"/dev", rfs, "/dev", "none", syscall.MS_BIND | syscall.MS_REC},
{"", "", "/proc", "none", syscall.MS_REC | syscall.MS_SHARED},
{"/proc", rfs, "/proc", "none", syscall.MS_BIND | syscall.MS_REC},
{"", "", "/sys", "none", syscall.MS_REC | syscall.MS_SHARED},
{"/sys", rfs, "/sys", "none", syscall.MS_BIND | syscall.MS_REC},
{"tmpfs", rfs, "/tmp", "tmpfs", 0},
},
argFlyMounts...,
)
for _, mount := range effectiveMounts {
var (
err error
hostPathInfo os.FileInfo
targetPathInfo os.FileInfo
)
if strings.HasPrefix(mount.HostPath, "/") {
if hostPathInfo, err = os.Stat(mount.HostPath); err != nil {
log.PrintE(fmt.Sprintf("stat of host directory %s", mount.HostPath), err)
return 1
}
} else {
hostPathInfo = nil
}
absTargetPath := filepath.Join(mount.TargetPrefixPath, mount.RelTargetPath)
if targetPathInfo, err = os.Stat(absTargetPath); err != nil && !os.IsNotExist(err) {
log.PrintE(fmt.Sprintf("stat of target directory %s", absTargetPath), err)
return 1
}
switch {
case targetPathInfo == nil:
absTargetPathParent, _ := filepath.Split(absTargetPath)
if err := os.MkdirAll(absTargetPathParent, 0700); err != nil {
log.PrintE(fmt.Sprintf("can't create directory %q", absTargetPath), err)
return 1
}
switch {
case hostPathInfo == nil || hostPathInfo.IsDir():
if err := os.Mkdir(absTargetPath, 0700); err != nil {
log.PrintE(fmt.Sprintf("can't create directory %q", absTargetPath), err)
return 1
}
case !hostPathInfo.IsDir():
file, err := os.OpenFile(absTargetPath, os.O_CREATE, 0700)
if err != nil {
log.PrintE(fmt.Sprintf("can't create file %q", absTargetPath), err)
return 1
}
file.Close()
}
case hostPathInfo != nil:
switch {
case hostPathInfo.IsDir() && !targetPathInfo.IsDir():
log.Printf("can't mount because %q is a directory while %q is not", mount.HostPath, absTargetPath)
return 1
case !hostPathInfo.IsDir() && targetPathInfo.IsDir():
log.Printf("can't mount because %q is not a directory while %q is", mount.HostPath, absTargetPath)
return 1
}
}
if err := syscall.Mount(mount.HostPath, absTargetPath, mount.Fs, mount.Flags, ""); err != nil {
log.PrintE(fmt.Sprintf("can't mount %q on %q with flags %v", mount.HostPath, absTargetPath, mount.Flags), err)
return 1
}
}
if err = stage1common.WritePpid(os.Getpid()); err != nil {
log.Error(err)
return 4
}
diag.Printf("chroot to %q", rfs)
if err := syscall.Chroot(rfs); err != nil {
log.PrintE("can't chroot", err)
return 1
}
if err := os.Chdir("/"); err != nil {
log.PrintE("can't change to root new directory", err)
return 1
}
diag.Printf("execing %q in %q", args, rfs)
err = stage1common.WithClearedCloExec(lfd, func() error {
return syscall.Exec(args[0], args, env)
})
if err != nil {
log.PrintE(fmt.Sprintf("can't execute %q", args[0]), err)
return 7
}
return 0
}